What actually changed
Microsoft's Advisor performance-recommendations reference now has a dedicated HorizonDB section. Each entry is tied to the resource type microsoft.horizondb/clusters and has its own recommendation ID. That makes this more than a generic "we support HorizonDB" line. Advisor is evaluating cluster settings and state and returning service-specific advice.
This integration is separate from the HorizonDB product launch. It is also not a brand-new Advisor interface. Advisor is described as a digital cloud assistant that helps you follow best practices to optimize your Azure deployments. You can see its recommendations as Owner, Contributor, or Reader of a subscription, resource group, or resource.
The product context
The timing fits Microsoft's Ignite 2026 messaging. In its November 17 Azure blog, Arun Ulag described Azure HorizonDB as now generally available. He called it a fully managed, PostgreSQL-compatible database with millisecond latency and built-in vector indexing. That post covers the service's GA status. It does not describe the Advisor integration.
One caution on status. Some earlier material still labels HorizonDB as a preview. For example, a Microsoft Tech Community roundup from earlier this year listed it under "Preview". The same-day Azure blog and the Azure Updates entry are the current statements. Treat older "preview" labels as stale.
The recommendations Advisor can raise
The performance reference lists these HorizonDB checks. Impact ratings are Microsoft's own.
| Recommendation | What triggers it | Suggested action | Impact |
|---|---|---|---|
| log_duration enabled | Logging adds overhead on every query | Set to OFF | Medium |
| log_error_verbosity = VERBOSE | Error logging more detailed than needed | Set to DEFAULT | Medium |
| log_min_duration_statement | Value under 60,000 ms (1 minute) | Raise the threshold | High |
| log_statement = ALL | Every statement logged | Use NONE or DDL | Medium |
| Autovacuum off | Autovacuum OFF and bloat ratio above 50% | Turn autovacuum on | High |
| log_statement_stats enabled | Adds logging overhead, can reduce throughput | Set to OFF | High |
| Table stats reset | Statistics were reset | Run ANALYZE | High |
| Approaching wraparound | Past the 50% wraparound limit, over 1 billion transactions | Follow the emergency autovacuum and wraparound guidance | High |
For the slow-query threshold, Microsoft says raising it reduces logging overhead, keeps logs smaller and helps you focus on the longest-running statements. For stats resets, it says to run ANALYZE, which collects statistics that help the PostgreSQL optimizer choose the best execution paths.
How to read these recommendations
Most of the list is about logging, which is a classic trade-off. Verbose logging helps when you are chasing a problem and costs you when you are not. A recommendation to turn off log_statement = ALL is sensible for steady-state production. It is a bad idea if you are in the middle of an audit or an incident investigation. Check why a setting was enabled before you change it.
Two items deserve different treatment:
- Wraparound: Transaction-ID wraparound is a PostgreSQL maintenance risk, not a tuning nicety. Microsoft's own thresholds are crossing 50% and over a billion transactions. Treat that alert as a priority.
- Autovacuum off with high bloat: This combination is more urgent than a logging tweak. It points to a maintenance setting that is actively hurting the cluster.
These are recommendations only. The documentation does not say Advisor changes any settings on its own. It also does not quantify the expected gains. I found nothing on regional rollout timing, eligibility or notification behavior, so I am not claiming any.
Practical steps for admins
- Open Azure Advisor in the Azure portal and go to the Performance recommendations.
- Filter by resource type to isolate HorizonDB clusters.
- Sort out the high-impact items first. Those are the slow-query threshold, autovacuum, statement stats, stats reset and wraparound.
- For each logging recommendation, confirm whether someone enabled the setting deliberately for diagnostics or compliance before you change it.
- Make the change through your normal change process and re-check Advisor afterward.
Advisor also has a REST API for listing recommendations. Teams that already pull Advisor data into reporting or ticketing can include HorizonDB through that route. I did not find HorizonDB-specific API guidance, so confirm the behavior in your own environment.
Bottom line
This is a modest but useful update. A new managed database is much easier to run when the platform itself flags the usual PostgreSQL mistakes: noisy logging, neglected autovacuum, stale statistics and wraparound risk. Admins evaluating HorizonDB should treat the Advisor list as a quick operational baseline. Because each recommendation is conditional, do not read an empty Advisor list as proof that a cluster is healthy.
References
- (Launched) Generally Available: Azure HorizonDB now integrates with Azure Advisor Azure Updates · 2026-11-17T20:00:32Z
- Your Privacy Choices Opt-Out Icon azure.microsoft.com
- Performance recommendations - Azure Advisor learn.microsoft.com